China a hub for deepfakes, approves 487 new algorithms from Baidu, Alibaba, Tencent

China has taken a significant step in regulating and advancing its artificial intelligence (AI) capabilities, particularly in deepfake technologies. The Cyberspace Administration of China (CAC) has approved 487 new AI algorithms, a move that reflects the country’s commitment to controlling and developing AI innovations.

These approved algorithms cover a range of applications from domestic tech giants like Baidu, Alibaba, and Tencent, as well as foreign firms including Hewlett-Packard (HP).

The latest approval list from the CAC is the second-largest since the deepfake regulation came into effect in January 2023. The newly approved algorithms include Baidu’s portrait image diffusion generator, which is integrated with its cloud gallery app Yike, and Tencent’s search algorithm used in WeChat. Alibaba has registered an algorithm for document creation in its enterprise collaboration tool DingTalk and a multimodal algorithm developed by its research arm, Damo Academy, for generating images and videos for corporate clients.

These efforts are part of the compliance with the Administrative Provisions on Deep Synthesis for Internet Information Services.

Chinese authorities heavily regulate and govern how deepfakes are used in the country. Failure to register these algorithms with the CAC could result in their removal from domestic app stores.

The CAC has been regularly publishing white lists of approved algorithms, with the latest list marking the seventh since the regulation was introduced. The largest list to date was published in June, featuring 492 algorithms.

Previous lists have included 394 algorithms in April and 266 in February. The first list in June 2023 had only 41 algorithms, indicating a rapid increase in approvals as the regulatory process has become more streamlined.

Cai Peng, a partner at Beijing-based Zhong Lun Law Firm, noted that the growing size of the lists indicates a clearer workflow between the regulator and applicants. The approximately two-month process involves filing and refining application documents based on the CAC’s requirements, suggesting that the regulator has established a more efficient system for handling applications.

The latest batch of approved algorithms includes several notable tools. Douyin, the Chinese version of TikTok developed by ByteDance, registered a healthcare knowledge algorithm. Microsoft’s AI spin-off Xiaoice had a music generator approved, and NetEase’s digital version of the popular party game Werewolf featured a character dialogue generator.

Several foreign brands also made the list. HP’s computer assistant algorithm was approved, alongside a text generator by Yum China, a spin-off of American fast-food giant Yum! Brands, used for dine-in and delivery services at KFC and Pizza Hut.

China is emerging as a global leader in AI regulation, requiring developers to register related technologies with the government. The country has mandated that generative AI (GenAI) models must be registered with the CAC before being made publicly available, a regulation that took effect in August 2023. To date, the CAC has released two lists, approving a total of 188 GenAI models.

China’s proactive stance in regulating AI technologies aims to ensure that innovations are developed and used responsibly, balancing advancement with control. The rapid approval of new algorithms demonstrates the country’s commitment to fostering a robust and secure AI ecosystem, setting a precedent for global AI governance.